MLV App 1.14 - All in one MLV Video Post Processing App [Windows, Mac and Linux]

[togg]

I have been experimenting quite a lot with processing Dual ISO footage on the MLV App V1.0 lately and have to say that it does an excellent job with converting, correcting and saving such files to ProRes.  I love the new sliders and the denoise function!  Rendering speed  is also faster compared to version 0.17.  Fantastic software, providing stunning results!  Thank you so much, guys!

I was wondering, does dualiso export is only to compressed format? Or can you export dualiso to dngs?

[IDA_ML]

Haven't tried that yet but it should work.  Why don't you try it by yourself?  I am away from my camera now.

[togg]

@togg: thx! I will download and analyse your clip later.
The speed point: v1.0 has many new parameters, so it might not feel faster, because they have to be calculated as well (if you use them). The new multithreading affects mostly RAW Corrections (e.g. try out chroma smooth 5x5). If only preview is faster: maybe you chose a faster debayer for preview than for export. Or do you export cdng? Then you won't feel much - exept RAW Corrections are enabled.Could you please explain - I don't understand that point. All options are saved into the session and should be the same after opening again. Just use Open and Save.

Ook so now I see. I do indeed use dng export, so as far as I understand for me multithread is more of an issue than anything, since it doesn't speed up export but use vastly more resources, ence heating the laptop for more hours etc I'll probably keep v0.17 running for longer export.

Yes, I can see now how options work with the session. Maybe what I think would be more usefull is the ability to set default settings for any new session, I never save one. I've tried to create a dummy session to reuse but it doesn't allow it to be empty, and if it can't find the old files it will revert to default. This is a problem because from time to time you can forget to reset things like vertical stripes etc.

Haven't tried that yet but it should work.  Why don't you try it by yourself?  I am away from my camera now.

I'm away from mine too, I'll get back to it in one week and test it

[masc]

Ook so now I see. I do indeed use dng export, so as far as I understand for me multithread is more of an issue than anything, since it doesn't speed up export but use vastly more resources, ence heating the laptop for more hours etc I'll probably keep v0.17 running for longer export.

Multithreading also speeds up export, also for dng, depending on the settings you did.

Yes, I can see now how options work with the session. Maybe what I think would be more usefull is the ability to set default settings for any new session, I never save one. I've tried to create a dummy session to reuse but it doesn't allow it to be empty, and if it can't find the old files it will revert to default. This is a problem because from time to time you can forget to reset things like vertical stripes etc.

Ok. Therefor you can save your standard receipts (Export Receipt) and load them (Import Receipt) onto every clip you wish. Or you load it just on one, Ctrl+C, Ctrl+A, Ctrl+V. Now your standard settings are on every clip.

For the sync: yes, I also see 1 to 1.5 frames difference. So this is the same like in my clip. I am not sure from where the problem comes. If we sync it in MLV App manually, the first video frame would be deleted or the first frame would have no audio. What we see seems to be like that in the MLV file.

[togg]

Multithreading also speeds up export, also for dng, depending on the settings you did.
Ok. Therefor you can save your standard receipts (Export Receipt) and load them (Import Receipt) onto every clip you wish. Or you load it just on one, Ctrl+C, Ctrl+A, Ctrl+V. Now your standard settings are on every clip.

For the sync: yes, I also see 1 to 1.5 frames difference. So this is the same like in my clip. I am not sure from where the problem comes. If we sync it in MLV App manually, the first video frame would be deleted or the first frame would have no audio. What we see seems to be like that in the MLV file.

Yes, clearly a module issue. I don't know if it is fixed or what in MLV lite, but imho the cleaner solution would be to delete the first 2(ish) frames all together.

Curious to know what other people results are with exported dngs, someone is exporting with only vertical stripe and bad pixel corrections applied?

[Danne]

Later versions of mlv_lite has adressed audio sync issues know to sometimes cause issues with mlv_rec. Bouncyball, like I do in Switch, I believe, is syncing audio by checking offset metadata in the mlv file. Did you record the unsyced audio with mlv_rec or mlv_lite? If yes then you know what to do.

[togg]

Later versions of mlv_lite has adressed audio sync issues know to sometimes cause issues with mlv_rec. Bouncyball, like I do in Switch, I believe, is syncing audio by checking offset metadata in the mlv file. Did you record the unsyced audio with mlv_rec or mlv_lite? If yes then you know what to do.

Yea, mlv_rec. In a week I'll have my camera back and test mlv_lite, do you feel it is a stable as mlv_rec?

[Danne]

More stable when it comes to sound. g3gg0 incorporated mlv_snd to mlv_lite and a1ex adressed sync issues to some sort of minimum. That together with offset syncing with post processong apps should make it more or less syncproof.
Check out this branch with sound:
crop_rec_4k_mlv_snd

[togg]

More stable when it comes to sound. g3gg0 incorporated mlv_snd to mlv_lite and a1ex adressed sync issues to some sort of minimum. That together with offset syncing with post processong apps should make it more or less syncproof.
Check out this branch with sound:
crop_rec_4k_mlv_snd

Ok nice! I'll do it very soon! I'll also have fun with higher fps and lossless compression. A lot of things to test before next project.

[bouncyball]

Yes, check in in any NLE. It is out of sync by 2 frames.

Got it.

Hint: Latest crop_rec_4k_mlv_snd branch has experimental audio setting (when mlv_audio.mo is loaded).
Go to:

Code: [Select]

ML Menu/Movie/RAW video/Sound recording/Audio delay
And try to experiment with that value (by default it is 1)

Later versions of mlv_lite has adressed audio sync issues know to sometimes cause issues with mlv_rec. Bouncyball, like I do in Switch, I believe, is syncing audio by checking offset metadata in the mlv file.

Sync is performed according to the 1st video and audio frame timestamp difference in the MLV. For the sync base 1st video frame timestamp is taken, audio can have earlier or later timestamp so positive or negative offset is calculated and used for cutting audio start or adding some silence at the beginnig, also audio start aligned to 2 audio sample boundary to prevent stereo channel swapping, the end size of the audio matched to the video length (for current FPS) and aligned to the size of audio block multiples (usually 4KB for 16bit/48khz) specific to current sample rate and bit depth (PCM audio WAV specification). That block size will be written into WAV header. The things get more complicated when MLV cutting is done but AFAIK algorithm we used worked good in any case tested.

Regards
BB

[masc]

I love the new sliders and the denoise function!

Thanks IDA_ML for all your nice words. Just one question: you tried out denoise and it was working for you? What OS did you use? Bouncyball tried it on Linux and it did nothing for him - that was the main problem, why it is only in the repos, but not in v1.0. For me it was working on all 3 OS.

[togg]

Got it.

Hint: Latest crop_rec_4k_mlv_snd branch has experimental audio setting (when mlv_audio.mo is loaded).
Go to:

Code: [Select]

ML Menu/Movie/RAW video/Sound recording/Audio delay
And try to experiment with that value (by default it is 1)
Sync is performed according to the 1st video and audio frame timestamp difference in the MLV. For the sync base 1st video frame timestamp is taken, audio can have earlier or later timestamp so positive or negative offset is calculated and used for cutting audio start or adding some silence at the beginnig, also audio start aligned to 2 audio sample boundary to prevent stereo channel swapping, the end size of the audio matched to the video length (for current FPS) and aligned to the size of audio block multiples (usually 4KB for 16bit/48khz) specific to current sample rate and bit depth (PCM audio WAV specification). That block size will be written into WAV header. The things get more complicated when MLV cutting is done but AFAIK algorithm we used worked good in any case tested.

Regards
BB

Perfect then, I'll be able to fine tune it until it works.

[bouncyball]

Bouncyball tried it on Linux and it did nothing for him

I have to apologize. Tested today and it indeed works with latest version. Blurs the image trying to eliminate luma grain not chroma noise (which in turn can be got rid of by chroma blur option).

[masc]

I have to apologize. Tested today and it indeed works with latest version. Blurs the image trying to eliminate luma grain not chroma noise (which in turn can be got rid of by chroma blur option).

Okay, thx, good to know. Yes, it is a very very simple median denoise filter. It blurs the noise, but edges only a little. For a little noise it might be enough, for ISO >1600 it isn't. But it is way faster than the other better algorithm I tried out ( 2min per frame?! )...

[bouncyball]

But it is way faster than the other better algorithm I tried out ( 2min per frame?! )...

[zalbnrum]

@masc

Hope I won't sound stupid again; is there a reason, that Cineform RAW is not included as an option of export? And those are unified files, not separate frames as in cDNG's? Lossles 1:4 > 1:10 cempressed 12bit RAW? I found there used to be Cineform RAW converter in early days of ML RAW (2014ish).

If it is so I think it would be really useful to implement that, if possible.

Thank you.

Link: https://gopro.github.io/cineform-sdk/

[IDA_ML]

Thanks IDA_ML for all your nice words. Just one question: you tried out denoise and it was working for you? What OS did you use? Bouncyball tried it on Linux and it did nothing for him - that was the main problem, why it is only in the repos, but not in v1.0. For me it was working on all 3 OS.

Masc,

I tried it on Win7x64 on my very old laptop and it works.  Denoising just takes longer.  I also use it in combination with the Chroma separation function at a Blur radius setting of 6 or 7.  This totally kills the color noise while preserving fine detail but it takes even longer to render.  Overall, I like the results from these too combinations very much.  In my opinion, this is a very powerful software. 

[Enrico]

Hi everyone, I've just a question, where could I find a well explained tutorial about the workflow of raw dual iso clips? I use to work with MLV App but I think this time i will need something more

[masc]

The only difference in the workflow is clicking the button dual iso. As far as I know there is no tutorial.

[togg]

I forgot to ask, regarding MLV lossless compression in the test branch that I'm going to test in a few days. It basically means that dng export will match the size if I use it right? They use the same compression? How about the stability? Is there any downside, even in post, if you capture compressed mlv? I'm only interested in 14bit.

[bouncyball]

Is there any downside, even in post, if you capture compressed mlv? I'm only interested in 14bit.

Believe me, there are only benefits

[masc]

The only downside should be: it is slightly slower in playback and rendering, because it has to be uncompressed.

[Danne]

There is also the possibility to uncompress the 14bit files if you want that...

[togg]

Believe me, there are only benefits

perfect!

The only downside should be: it is slightly slower in playback and rendering, because it has to be uncompressed.

That could be an issue, I'm already bothered by the fact that every time that you want to play a new clip, especially if it is a long one, you have to wait such a longtime. It doesn't work very well for dailies.

There is also the possibility to uncompress the 14bit files if you want that...

mmm I would probably avoid an extra step like this, since I always need to export dng for resolve.

[theBilalFakhouri]

Why not to add Bilinear Cached preview for Real-time or mostly Real-Time ?